I’d been distracted from the water drop project for a while but, after handing in a magazine story plus images on Friday I’ve had time to give it a good burst of energy. I had some success, finally managing to photograph some actual collisions. Here’s one:
In my last post I described my setup. I’d been having trouble getting consistent results from the cheap solenoid valve I was using. I decided that I needed a better way of studying the behaviour of the valve. I came across “Droplet”, a wonderful little piece of software for the Arduino (the tiny micro computer that I use for controlling the valve), written by a photographer called Stefan Brenner. (He is a real master of ‘liquid art’ photography. Check out his portfolio here. )
With ‘Droplet’, I can connect my laptop to the Arduino and use it to control a whole lot of devices: solenoid valves, flashes, cameras. So for instance in that photograph above the following things happened:
- the camera shutter was triggered, set for 1/2 sec exposure
- the solenoid was opened for 15 milliseconds (ms), then closed again. That let go one water droplet.
- 69 ms after the first opening, the valve was opened again for 15 ms, to allow a second water drop to fall.
- 420 ms after that first opening, the flash went off (for about 1/30,000 of a second). Well, 3 flashguns actually.
- 30 ms after that the camera shutter closed
Using ‘Droplet’ I was able to check how different variables affected the quality and consistency of the water drops generated by the valve. It didn’t perform particularly well, but I did manage to get some interesting shots (and some good data).
Easiest was the classic single droplet shot:
This next one is either the droplet detached from the upwards moving jet, or the second droplet coming down at speed. I’m not yet sure which.
Under some conditions the discharge from the valve is really ugly: several droplets of varying size. This plays havoc with the shots. In this next shot a small second droplet has landed off to one side, while a large, third droplet has arrived early and wiped out the column before it had a chance to form properly.
The alignment of the valve nozzle appears to be quite important. The droplets emerge with a slight velocity and, if there’s a sideways component it’s sufficient to make the second droplet hit to one side of the jet. That’s what happened in that first picture at the beginning of this post.
By altering the camera angle, I could get a better view of a very similar collision:
Timing is absolutely critical. 1 millisecond can make quite a difference to the result! Fortunately, the Arduinos are fast. Once I had found the right timings and conditions, I was able to get collisions like these about one in every two or three attempts.
I’m now at the stage of deciding whether to continue with this valve, or replace it with a better one. The model that Stefan uses is not readily available here in New Zealand (and expensive). Some people use as many as three valves, although I think it will be quite a challenge to get all those droplets to go exactly where I want them, when I want them. I’m toying with the idea of ordering some from China. I have been impressed with the build quality of some of the electronic gear I have been ordering from China, and their customer service, so I’m tempted to try that source.
Of course, none of these images is particularly impressive, when compared with what has been achieved by many others. But it’s progress! Once I get some repeatability, then I can start playing around with different fluids, lighting setups, numbers of droplets and valves, etc.
Just to be clear: all of these images were taken without using any of the HiViz gear described in my first post on this topic (except for an opto-isolator to protect the camera). Instead of using photo sensors and analogue delay units to trigger a flash, I was relying on precise timings. These had to be worked out by trial and error, made a whole lot easier by Stefan Brenner’s program.
2 April: a postscript. I’m expecting some new valves any day now, but here’s where I”ve got to with the old one: